nvidia eyes tsmc's a16: a power play for ai dominance
Hsinchu, Monday, 15 September 2025.
nvidia may be the first to adopt tsmc’s a16 process. This could cement its lead in the ai chip market in 2025. For the first time an AI application would dominate tsmc’s most advanced process. The a16 process promises to boost the computing power and efficiency of nvidia’s gpus. This widens the gap with competitors like amd. This move marks a shift for nvidia, which has relied on older processes. The a16 chip is said to increase speed 8-10% at the same voltage and reduce power 15-20% at the same speed.
a16 process details
TSMC’s A16 process is a further refinement of its N2 technology [1]. It incorporates advanced transistor structures, high-precision lithography, and innovative materials [1]. These enhancements aim to improve energy efficiency, transistor density, and overall performance [1]. Super Power Rail (SPR) technology is a key component, relocating power lines to the wafer’s backside [1]. This frees up space on the front side for signal lines, boosting logic density and performance, while also reducing voltage drop and improving power efficiency [1].
nvidia’s strategic shift
Nvidia’s potential adoption of the A16 process marks a significant change in strategy [2]. Historically, Nvidia has not been among the first to utilize TSMC’s most advanced nodes [2]. Instead, companies like Apple, MediaTek, and Qualcomm have typically led the way [2]. This time, Nvidia is prioritizing cutting-edge process technology to enhance its AI architectures [2]. Nvidia last used TSMC’s most advanced process technology around 2004 with its 110 nm chips [4]. By embracing A16, Nvidia signals its commitment to leveraging manufacturing advancements to maintain its competitive edge [1].
market and stock impact
Analysts predict that the A16 process will be a game-changer for the AI chip market [3]. It will enable the creation of more powerful and efficient AI accelerators [3]. TSMC’s revenue is expected to increase as mainstream AI hardware manufacturers adopt its high-end nodes [2]. News of Nvidia’s potential adoption of the A16 process could positively impact its stock [GPT]. This reflects sustained technological leadership and the potential for increased revenue and market share [GPT]. TSMC’s revenue for the first eight months of 2025 reached 335.772 billion New Taiwan dollars, a 3.9 % increase month-over-month and a 33.8 % increase year-over-year [3].
future applications and competition
Nvidia may use the A16 process in its next-generation AI training chips [1]. This could include successors to the Blackwell architecture or a new GPU architecture codenamed “Vera” [1]. The A16 process is expected to enter trial production in the second half of 2026, with mass production in 2027 [1]. The high density and low power consumption of A16 will allow Nvidia to integrate more cores and cache on a single chip [1]. This reduces energy costs for data centers [1]. AMD is also expected to use TSMC’s 2nm process for its high-performance computing chips [3].